> ## Documentation Index
> Fetch the complete documentation index at: https://docs.onerun.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Create a Simulation

> Step-by-step guide to setting up and running your first agent evaluation simulation

This guide walks you through creating and running a complete simulation to evaluate your AI agent's performance. You'll learn how to set up projects, agents, objectives, and launch simulations with generated personas.

## Prerequisites

* OneRun platform running (see [Quick Start](/quickstart) for setup)
* Access to OneRun web interface
* Basic understanding of your agent's capabilities

## Step 1: Create a Project

Projects organize all your evaluation work and provide isolation between different agent testing initiatives.

1. **Open OneRun** in your browser
2. **Sign in** to your account
3. **Click "New Project"** on the dashboard
4. **Enter project name**: Choose a descriptive name (e.g., "Customer Support Production" or "Marketing Team Bots")
5. **Save the project** - you'll be redirected to the project dashboard

<Tip>
  Use project names that identify the business function or environment rather than specific agent versions. This helps organize multiple agents and evaluations within the same business context.
</Tip>

## Step 2: Define Your Agent

Agents represent the AI system you want to evaluate. The agent configuration helps OneRun understand what your system does and generates appropriate test scenarios.

1. **Navigate to "Agents"** in your project

2. **Click "Create Agent"**

3. **Configure agent details**:
   * **Name**: Your agent's name (e.g., "Support Bot v2.1")
   * **Description**: Detailed description of what your agent does, its capabilities, and role (e.g., "Friendly customer service agent that handles billing inquiries, processes returns, and provides product information")

4. **Save the agent** - note the Agent ID from the details page (needed for your worker)

<Note>
  The agent description is crucial as OneRun uses it to generate realistic personas and conversation scenarios. Be specific about your agent's capabilities and limitations.
</Note>

## Step 3: Set Evaluation Objectives

Objectives define what "success" looks like for your agent. They provide the scoring criteria that OneRun uses to evaluate conversation quality.

1. **Go to "Objectives"** in your project

2. **Click "New Objective"**

3. **Define success criteria**:
   * **Name**: Clear objective name (e.g., "Customer Satisfaction")
   * **Criteria**: Detailed evaluation guidelines (e.g., "Evaluate how satisfied the customer feels with the interaction. Score 1.0 for highly satisfied customers who express gratitude. Score 0.5-0.8 for neutral interactions where basic needs are met. Score 0.0-0.4 for frustrated customers or unresolved issues.")

4. **Add multiple objectives** for comprehensive evaluation:
   * **Response Accuracy**: How factually correct are the agent's responses?
   * **Customer Satisfaction**: How satisfied is the customer with the interaction?
   * **Professional Communication**: Does the agent maintain appropriate tone and language?

<Warning>
  Create specific, measurable objectives. Vague objectives like "good response" make it difficult to evaluate and improve your agent consistently.
</Warning>

## Step 4: Create a Simulation

Simulations bring everything together - they generate personas, orchestrate conversations, and evaluate results against your objectives.

1. **Navigate to "Simulations"**
2. **Click "New Simulation"**
3. **Configure simulation parameters**:
   * **Name**: Descriptive name for this test run
   * **Agent**: Select the agent you created
   * **Objectives**: Choose which objectives to evaluate
   * **Scenario Description**: Describe the situations you want to test (e.g., "Customers with billing issues, product returns, and general inquiries")
   * **Number of Conversations**: Start with 10-20 for initial testing
   * **Max Turns per Conversation**: Set appropriate limits (3-5 for simple tasks, 10+ for complex scenarios)

## Step 5: Launch the Simulation

With everything configured, you're ready to run your first evaluation.

1. **Review simulation settings** - make sure everything looks correct
2. **Click "Start Simulation"** - OneRun will begin generating personas and scenarios
3. **Monitor progress** - you'll see conversations being created and assigned
4. **Ensure your worker is running** - without a worker, conversations won't proceed

<Steps>
  <Step title="Persona Generation">
    OneRun creates diverse personas based on your agent description and scenario
  </Step>

  <Step title="Conversation Assignment">
    Each persona gets assigned to a conversation with your agent
  </Step>

  <Step title="Worker Processing">
    Your worker polls for conversations and handles the agent logic
  </Step>

  <Step title="Evaluation">
    Completed conversations are automatically scored against your objectives
  </Step>
</Steps>

## Next Steps

After creating your first simulation:

* **Scale Up**: Run larger simulations with 50+ conversations for statistical significance
* **Compare Versions**: Use simulations to A/B test agent improvements
* **Automate Evaluation**: Integrate OneRun into your development pipeline
* **Share Results**: Use reports to communicate agent performance to stakeholders

<Note>
  Effective agent evaluation is iterative. Use each simulation to learn something new about your agent's capabilities and limitations.
</Note>

Ready to implement your agent? Check out [Connect an Agent](/guides/connect-an-agent) for detailed worker implementation.